The Motivations for the Adoption of Management Innovation by Local Governments and its Performance Effects

: This article analyses the economic, political, and institutional antecedents and performance effects of the adoption of shared Senior Management Teams (SMTs)—a management innovation (MI) that occurs when a team of senior managers oversees two or more public organizations. Findings from statistical analysis of 201 English local governments and interviews with organizational leaders reveal that shared SMTs are adopted to develop organizational capacity in resource-challenged, politically risk-averse governments, and in response to coercive and mimetic institutional pressures. Importantly, sharing SMTs may reduce rather than enhance efficiency and effectiveness due to redundancy costs and the political transaction costs associated with diverting resources away from a high-performing partner to support their lower-performing counterpart.

P ublic organizations are increasingly called upon to design and implement innovative management practices to become more effective and efficient (Brown, Osborne, and Walker 2016;Kim and Warner 2016). However, despite the growing emphasis on management innovation (MI) in the public sector, evidence on the economic, political, and institutional factors that lie behind its adoption is only slowly emerging (Damanpour and Aravind 2011;Singla, Stritch, and Feeney 2018). More significantly, very little is known about whether MIs generate the anticipated organizational improvements (De Vries, Bekkers, and Tummers 2015;Walker, Chen, and Aravind 2015). In this article, we investigate the antecedents and performance effects of MI in public service organizations. Specifically, we examine shared Senior Management Teams (SMTs) in English local governments-a MI that occurs when two or more local governments formally agree that their service delivery will be overseen by a single group of senior managers.
Broadly defined, MI involves the introduction of a new structure, process, system, program, or practice in an organization or its units which changes how managers manage (Birkinshaw, Hamel, and Mol 2008). MIs are therefore distinct from other innovation types (e.g. service, partnership) that entail material changes to what an organization provides to its external stakeholders (Damanpour and Aravind 2011). Because the adoption of MIs may result in the redesign and redevelopment of management systems and processes, it is often radical rather than incremental in orientation (Walker, Damanpour, and Devece 2011). MI may therefore be particularly challenging for public organizations, where changes to established rules and routines can generate high economic and political transaction costs (Rodrigues, Tavares, and Araújo 2012).
To understand why local governments choose to share SMTs, we investigate rational and institutional perspectives on the adoption of MI (Birkinshaw, Hamel, and Mol 2008), developing and testing theory relating to the economic, political, and institutional antecedents likely to influence local government decision-makers. From a rational perspective, sharing a single SMT may be attractive to smaller organizations seeking to generate scale economies and build additional management capacity, especially in response to fiscal stress (Kim and Warner 2016). At the same time, fierce electoral competition and short electoral The Motivations for the Adoption of Management Innovation by Local Governments and its Performance Effects cycles may make politicians unwilling to introduce potentially controversial policy changes (Berry and Berry 1992), while rightwing governments may believe it is in their electoral interest to adopt disruptive MIs (Bel and Fageda 2017). From an institutional perspective, coercive pressures from regulatory bodies, and higher levels of government could prompt local governments to search for new ways of doing things (Ashworth, Boyne, and Delbridge 2009). At the same time, mimetic pressures from nearby adopters of innovation are also likely to be influential .
Whether rational or institutional motivations behind the adoption of MI predominate, public organizations are unlikely to share SMTs unless they anticipate a performance pay-off. Indeed, although empirical evidence on the performance effects of MI is scant (Walker, Chen, and Aravind 2015), reforms intended to promote MI remain a popular prescription for improving public service performance across the globe (OECD 2017). Nevertheless, theory and evidence on the transaction costs associated with shared service provision suggests that sharing local governments may experience few performance gains and could even suffer losses McQuestin and Drew 2019).
To analyze the antecedents and performance effects of the adoption of shared SMTs by English local governments, we use both quantitative and qualitative data. Following best practice in recent quantitative research on public sector innovation (e.g. Yi and Chen 2019;Zhang and Zhu 2020), longitudinal data for the period 2006-2014 are drawn from well-established secondary sources. Findings from statistical analysis are then complemented with interviews with a wide range of local government chief executive officers (CEOs), senior managers, and politicians, as per qualitative studies of local government innovation (e.g. Carassus, Favoureau, and Gardey 2013). In the conclusions, we elaborate on the implications of the findings.

Antecedents of Shared Senior Management Teams
Local governments across the globe have undertaken an array of structural reforms and innovations in response to contemporary demands for more cost-effective public service provision (Dollery et al. 2020). In the United States and Western Europe, such reforms have often involved the consolidation and amalgamation of smaller units of government into larger ones (Faulk and Hicks 2011;Kuhlmann and Bouckaert 2016), but, increasingly entail the development of collaborative structures intended to build capacity, realize scale economies, and enhance service integration across multiple jurisdictions (Tavares and Feiock 2018). Within this context, the sharing of services along with their functional management structures has become a common approach to resolving collective action dilemmas (Morse and Abernathy 2015). However, until recently, the joint management of entirely separate local government entities by a single SMT has been a comparatively neglected structural reform, even though the institutional requirements for its implementation may be similar to those needed for sharing services.
Shared SMTs occur when a team of senior managers oversees two or more public organizations. In England, local governments seeking to share SMTs can only do so once each of the sharing governments has voted to approve the new arrangements. 1 Critically, shared SMTs involve a high degree of organizational integration without the loss of organizational identity and sovereignty because the governments that share a SMT retain separate political oversight. Research suggests that inter-local co-operation has grown considerably in the United States (US) and many other countries during the past thirty years (Kwon and Feiock 2010;Bel, Mildred, and Warner 2015). In particular, empirical studies have identified a movement toward the sharing of services, such as IT, HR, and procurement, in the United States (Henderson 2014), Australia (Drew, McQuestin, and Dollery 2019), England , and Germany (Niehaves and Krause 2010). However, to date, little systematic empirical analysis has addressed the reasons for appointing a single SMT to run all of the services provided by separate local governments. 2

Economic Antecedents
Theories of MI in the public sector emphasize the importance of supply-side factors that reduce the economic transaction costs associated with making innovations work. In particular, organizational characteristics associated with the realization of scale economies, such as large size, high administrative intensity, and additional revenues, are assumed to be essential in reducing the time and money needed to search for new ways of working (Walker 2014). However, for a MI that brings similar organizations together in pursuit of greater capacity, such as sharing a SMT, it may be that small organization size, low administrative intensity, and reduced revenues become demand-side factors prompting collaborative efforts in order to reduce transaction costs (Shrestha and Feiock 2011).
Small organizations are flexible and adaptive to changing environmental circumstances, and can be adept at implementing new practices and routines (Damanpour 1992). However, they are also prone to higher economic transaction costs because they lack the capacity that enables larger organizations to exert greater control over the supply and use of their material resources (Nooteboom 1993). As a result, small organizations tend to feel the pressures of resource dependence more than large organizations (Pfeffer and Salancik 2003). This may be especially important for small units of government in England, where successive central governments have amalgamated municipalities and consolidated the local government system, partly in response to the perceived co-ordination challenges posed by fragmentation . Moreover, the realization of the benefits of administrative intensity is likely to increase with scale and scope (at least up to a point) (Rutherford 2016). Hence, the opportunity to share managerial resources across two or more public organizations creates the potential for organizations with small central bureaucracies to reduce economic transaction costs by generating administrative scale and scope economies. Given the wider context of collective action dilemmas confronted by local governments across Europe (Tavares and Feiock 2018), we therefore test two hypotheses that recognize the importance of economic transaction costs and resource dependence as a driver of the MI investigated: Hypothesis 1: Organizational size is negatively related to the adoption of a shared SMT.
Hypothesis 2: Administrative intensity is negatively related to the adoption of a shared SMT.
Changes in the financial resources available to public organizations are likely to affect attempts at innovation (Wolman 1986). In particular, when resources decline, public managers may seek to develop and implement innovations that reduce their resource commitments (van der Voet 2019). MIs that involve resourcesharing are therefore likely to be attractive to public organizations, such as English local governments, that face budget cuts, with limited autonomy to raise revenue, and growing demand for services they have a statutory duty to provide. Indeed, research suggests that poor fiscal health leads local governments to become more entrepreneurial (Kim and Warner 2016;Singla, Stritch, and Feeney 2018). Thus, our third hypothesis is: Hypothesis 3: Expenditure reductions are positively related to the adoption of a shared SMT.

Political Antecedents
Politics is also likely to influence the adoption of shared SMTs by local governments. In particular, the ideological commitments of the ruling party can explain policy choices (Besley and Coate 1997). In theory, right-wing parties cut costs and enhance efficiency, whereas left-wing parties favor state intervention to improve equity. For this reason, the adoption of MI aimed at resource-sharing seems more likely to occur in local governments controlled by right-wing parties (Bel and Fageda 2017). Most local governments in England are controlled either by the Conservative (right-wing) or Labour (left-wing) party. In general, the Conservatives favor low taxation and public service expenditure, whereas Labour usually champions public spending on services (Smith 2010). We therefore posit: Hypothesis 4: Right-wing party rule is positively related to the adoption of a shared SMT.
The policy adoption literature suggests that incumbent politicians introduce new policies and programs to broaden their electoral appeal (Somer-Topcu 2009). However, strong electoral competition may make them reluctant to introduce radical innovations that do not result in immediate benefits for key constituents. Indeed, policy design theory suggests that controversial policies or initiatives are less likely to be adopted where the risk of electoral failure is higher (Boushey 2016). As a result, local governments may only have the confidence to share SMTs when they are safe in the knowledge that they are unlikely to be voted out. For this reason, where the electoral cycle is short, the adoption of MI is likely to be less attractive for politicians, as there will be less time available for the learning required to iron out any implementation problems. Several studies suggest that the adoption of controversial policies, such as tax increases and school choice, often occurs just after an election has taken place (e.g. Berry and Berry 1992;Mintrom 1997). All of these arguments about political risk aversion seem likely to apply to MIs, as results may take time to emerge and there is a risk of failure. We therefore propose: Hypothesis 5: Electoral competition is negatively related to the adoption of a shared SMT.
Hypothesis 6: Time until next election is positively related to the adoption of a shared SMT.

Institutional Antecedents
The institutional perspective on MI focuses on external influences on organizations (Birkinshaw, Hamel, and Mol 2008). From this point of view, the pursuit of legitimacy within the institutional context is the primary objective for public (and private) organizations (Scott 2014), which, in turn, encourages conformity in the management practices that they adopt (Ammons and Roenigk 2015). DiMaggio and Powell (1983) describe this process as institutional isomorphism, and distinguish between coercive pressures that are the product of the regulatory environment in which organizations operate, mimetic pressures resulting from policy diffusion throughout organizational populations, and normative pressures caused by stakeholder expectations regarding appropriate organizational behavior.
Regulatory pressures have previously been shown to increase the likelihood of institutional isomorphism in the English local government system (Ashworth, Boyne, and Delbridge 2009). However, analysis of variations in the strength of coercive isomorphism in English local government is very difficult, because all local governments are subject to the same regulatory regime. One potentially promising approach to understand general coercive isomorphic pressures would be to investigate the distance between local governments and the center of state power within a country. England is regarded as being a highly centralized state, in which the central government is content to concede greater autonomy to organizations at the geographical periphery due to their perceived lack of importance (Ayers et al. 2018). In such circumstances, it seems prima facie likely that local governments close to the capital city will be subject to greater coercive pressures than those further away. Nevertheless, local governments that have better access to national capitals may have informational advantages over their more peripheral counterparts, and are thereby better able to lobby for their interests (Goldstein and You 2017;Greer and Sandford 2006). For that reason, we advance: Hypothesis 7: There is an inverted u-shaped relationship between proximity to central government and the adoption of a shared SMT.
During the 2000s, local governments in England had to demonstrate their legitimacy to escape the extension of political control from higher levels of government (Ashworth, Boyne, and Delbridge 2009). One way in which public organizations can build legitimacy and avert political intervention is to copy the innovations adopted by their neighbors. Such mimetic behavior is especially likely in a horizontal peer network, where performance comparison can highlight performance gaps that potentially threaten organizational legitimacy (Villadsen 2013). Mimicry of the management practices of their neighbors may be especially important for local governments, such as English district councils, that are under threat of amalgamation or abolition (Thurmaier and Wood 2016). In the face of strong institutional pressures, it seems probable that the decision to adopt MI for any given local government will be influenced by the choices made by its neighbors, and that this behavior is spatially dependent. Indeed, previous studies have highlighted spatial dependence in public service innovations (Rincke 2006). For this reason, we suggest: Hypothesis 8: Proximity to prior adopters is positively related to the adoption of a shared SMT.

Performance Effects of Shared Senior Management Teams
Drawing on the rational perspective on MI, it is possible to identify two key reasons for expecting positive performance impacts from sharing a SMT. Firstly, as discussed above, the adoption of a shared SMT may help to capture economies of scale. In particular, fixed administrative overheads can be spread across more units of output (Boyne 1995), and organizations sharing a SMT likely have greater purchasing power, enabling them to reduce supplier costs and recruit better personnel (Black, Noel, and Wang 1999). Secondly, theories of organizational learning indicate that when implementing MI, organizations can generate valuable new information about how to improve efficiency and effectiveness (Stata 1989). In the case of shared SMTs, organizations can potentially learn from each other's successes and failures in a more systematic way than would normally be possible (Lasker, Weiss, and Miller 2001).
For these reasons, one might anticipate that adoption of a shared SMT will be positively related to public service performance. However, institutional collective action theories point to the significant transaction costs associated with inter-municipal co-operation (Bel, Mildred, and Warner 2015;Tavares and Feiock 2018). Such costs arise from the need to monitor and manage new and unanticipated demands, information asymmetry, and potentially opportunistic behavior by partner organizations (Shrestha and Feiock 2011). Although some of the economic transaction costs associated with collaboration can be internalized by sharing a SMT, there will still be significant political transaction costs associated with negotiating and securing agreement between two or more "sovereign" political entities (Lubell et al. 2017;Rodrigues, Tavares, and Araújo 2012). Indeed, prior empirical research investigating the effects of shared local service production indicates that administrative efficiency does not improve  and could deteriorate (McQuestin and Drew 2019). Hence, we propose: Hypothesis 9: Adoption of a shared SMT will be negatively related to public service performance.

Unit of Analysis
Our units of analysis are the full population of 201 district councils in England: multi-purpose local governments operating within the lower level of the two-tier system serving nonmetropolitan areas. These councils provide mostly neighborhoodlevel public services, such as leisure centers, waste management, and residential planning. They serve small cities and towns, but are comparatively large by European standards (mean population of 101,234, with an average annual expenditure of £312 per capita). District councils are elected bodies usually with a cabinet system of political management made up of senior members of the ruling political party. The politicians implement national policy frameworks on the advice of professional local government managers led by a CEO and a team of appointed senior corporate and functional service directors.

Data
Our study combines quantitative and qualitative data to address the hypotheses from multiple perspectives and enrich our description and analysis of the antecedents and effects of shared SMTs. We apply a concurrent nested design that involves collecting quantitative and qualitative data and integrating the results in the analysis phase to deepen our interpretation of the phenomenon (Creswell and Plano Clark 2011). Quantitative data guides our research and the hypotheses testing, and qualitative data play a supporting role by deepening understanding of the motivations behind the decision to share a SMT and its perceived impacts. The interviews aimed at adding more nuanced findings to better explain, from the managers' and politicians' perspectives the reasons behind the decision to share a SMT, the circumstances that led each council to the decision and the perception of the impacts achieved so far.
Following the example of recent high-quality innovation research in management studies (e.g. McElheran 2015), political science (e.g. Desmarais, Harden, and Boehmke 2015), and public administration (e.g. Mallinson 2020), we draw on a range of highquality quantitative longitudinal data. The data were collected from the Ministry for Housing, Communities and Local Government, the Office of National Statistics, Sport England and the annual accounts of district councils. The dataset includes information from 2003 to 2014. Data on which district councils had adopted a shared SMT, and when, was gathered through documentary analysis of council reports, business plans, and minutes from council meetings. Our search revealed 37 district councils involved in 18 shared SMT arrangements by 2014. We subsequently contacted all of these councils to collect information on their approach to sharing a SMT. We found that the first shared SMT started in 2007.
As per Carassus, Favoureau, and Gardey's (2013) study of innovation in three French local governments, qualitative data were collected from a range of key actors. Specifically, twenty-nine semi-structured interviews with CEOs, senior managers, and politicians in a sample of nine councils that are involved in four shared SMT arrangements: three involved two councils, with one involving three councils. We selected the four sample cases of sharing from different parts of England to provide geographical variation, and focused on early adopters because of the information-rich perspective on the operation of this MI that they could offer. We interviewed key actors who had an important role in the adoption of the MI and/or were currently contributing to their implementation. The large majority of interviews were conducted face-to-face (the remainder were conducted by telephone) between the end of 2016 and the beginning of 2017. Interviews lasted 45-60 minutes and were semi-structured with a core set of questions exploring key themes (e.g. motivations behind the decision to share the SMT, contextual background, impacts achieved so far) combined with open questions to allow for the emergence of unanticipated insights (e.g. unforeseen outcomes such as greater organizational resilience). Interviews were audiotaped and transcribed before being analyzed by two of the authors using Atlas.ti. The main parts of the interviews' content were marked with a series of codes using the headings and sub-headings from the topic guide but also open codes (e.g. motivations: save money; antecedents: political control; impacts: more capacities, skills, and resilience, etc.). The full list of interviewees is shown in the Appendix (Table A1). The coding frame for our qualitative data analysis is available on request.
The Motivations for the Adoption of Management Innovation by Local Governments and its Performance Effects 5

Management Innovation
For our quantitative analysis, we investigate the adoption of a shared SMT using survival analysis. The dependent variable in this case is the hazard rate for switching from being a non-sharing council to one that shares its SMT, which can be characterized as the likelihood that this particular event will happen to a given organization at a particular time. The hazard of switching to shared management arrangements is a function of the length of time it takes for a council to adopt a shared SMT. District councils that survived the study period without sharing their SMT are assigned a value of 1 on a status variable. Councils that adopted a shared SMT received a 0 from the year in which they switched management model. 3 Public Service Performance. The measurement of performance in the public sector is complex, multidimensional, and shaped by the perspectives of multiple stakeholders (Boyne 2003). The complexity of the public service performance construct suggests an ideal analysis of the relationship between MI and performance would comprehensively describe the achievements of an organization across a range of performance dimensions using perceptual and archival data collected from internal and external stakeholders (Andersen, Boesen, and Pedersen 2016). We focus here on two aspects of local government performance for which high-quality archival data are available for the entire study period: public service efficiency and effectiveness. Both of these dimensions of performance are of major concern to two key stakeholders in local government: i) local citizens as taxpayers and service users; and, ii) the U.K. central government as the main source of funding for district councils.
Following the approach of local government economists (e.g. Narbón-Perpiñá, Balaguer-Coll, and Tortosa-Ausina 2019), we measure efficiency by applying Data Envelopment Analysis (DEA) techniques to a series of inputs and outputs collected for our pooled data set. We compute radial distance measures based on the Debreu-Farrell notion of efficiency and adopt an inputorientated DEA model because local government managers have greater control over the level of inputs than outputs. We also assume variable returns to scale (see, De Borger and Kerstens 1996; Narbón-Perpiñá, Balaguer-Coll, and Tortosa-Ausina 2019).
The selection of outputs for our DEA model is based on the key services provided by all English district councils. We, therefore, include the following indicators as outputs: (i) number of adults participating in regular, moderate intensity sport (leisure and culture services); (ii) tons of waste managed (environmental services); (iii) total CO 2 emissions managed (environmental services); (iv) number of planning applications received (planning services); (v) number of active businesses (administrative services); (vi) number of taxable households (administrative services); (vii) population (a proxy for all council services-De Borger and Kerstens 1996). Due to data availability, our DEA model comprises published information from 2006 to 14. Councils sharing a SMT may be able to produce a greater quantity of these outputs if they are able to share managerial expertise, but may potentially produce fewer of these outputs due to the political transaction costs associated with making joint management decisions for two sovereign bodies (see below).
To capture inputs, we aggregate local governments' expenditures on the key services of leisure and culture, environment, administration, and planning. The summary statistics for our DEA model are provided in Table 1.
To measure effectiveness, we draw upon publicly available performance indicators used by U.K. central government to monitor the achievements of district councils. To compare these across different service areas, we first inverted some (e.g. carbon dioxide emissions per capita) so high scores always denote high performance. We then took z-scores of each indicator and combined them to create an overall index of effectiveness for the years 2006-14. We specifically focused on key performance indicators for each of the main services provided by district councils (see Table 2). For example, rates of waste recycling were included within the index as a measure of the effectiveness of the waste management services provided by councils, and the rate of sports participation among adults to gauge the effectiveness of leisure services provision. For councils sharing a SMT, achievements on these performance indicators may depend upon the SMTs' ability to devote sufficient managerial attention and expertise to each of the sovereign entities for which they are responsible. This may be particularly challenging if one of the sharing councils is performing much worse than the other(s) (see more below).

Antecedents of Shared SMTs
We use eight variables to test our hypotheses on the antecedents of shared SMTs. First, organization size is measured as the population served by each district council-a commonly used indicator of local government size that can capture the full scope of governments' output (e.g. De Borger and Kerstens 1996). Second, administrative  intensity is measured as the percentage of each district council's overall expenditure that is spent on central administrative services (see Dixon and Elston 2020). Third, we measure expenditure reductions as the annual percentage change in total expenditure per capita for each council (for a similar approach see Jordan 2003). The polarity of the sign for this variable is reversed to capture spending reductions.
To analyze the political factors shaping adoption of a shared SMT, right-wing party rule is measured using a dichotomous variable coded 1 for councils controlled by the Conservative Party and 0 otherwise (see Dixon and Elston 2020). Next, electoral competition is measured by calculating the percentage difference of the share of the vote gained by the two parties winning the most votes in the most recent local election subtracted from 100 (see Pattie and Johnston 2005). Finally, the influence of time until next election is measured as the years until the next local election (see Berry and Berry 1992).
To investigate the institutional factors influencing shared SMT adoption, we utilize indicators of geographical proximity (see also Zhang and Zhu 2020). First, to capture coercive institutional pressures we calculate the distance from the headquarters of each district council to the offices of the Department for Communities and Local Government in London. 4 To test for a nonlinear relationship between proximity to central government and MI adoption, we include the base term and a squared version of this measure in our model. Second, to gauge mimetic institutional pressures, we compute the number of adjacent district councils that share a SMT using a row-normalized spatial contiguity matrix, a common approach in policy diffusion research (see Cook, An, and Favero 2019).
In addition to the independent variables, we included measures of innovative culture/previous innovative practices, financial status, and leadership that are commonly applied in high-quality public administration research (see Rho and Han 2020; Yi and Chen 2019; Zhang and Zhu 2020). First, to proxy for the potential influence of an innovative culture on the adoption of MI, we included an annual count of the companies operated by each district council, drawing on information in their annual accounts. Councils that make extensive use of municipal companies are regarded as more entrepreneurial than those that do not (Skelcher 2017). Second, we added a measure of the financial reserves per capita held by each council as a proxy for their overall financial condition (Jacob and Hendrick 2013). Third, we included a measure of the tenure (in years) of the political leader of each council to capture the extent to which more experienced leaders may be more likely to enact innovation (Korac, Saliterer, and Walker 2017). To access accurate public records of district political leaders' identities for the entire study period (2003-14), we submitted Freedom of Information requests to all the district councils in England. Analysis revealed that the tenure-MI relationship followed a nonlinear inverted u-shaped pattern (a phenomenon identified in some previous research, e.g. Miller 1991). In Table 4 we report the results including the squared and non-squared versions of the political leader tenure variable.

Performance Effects of Shared SMTs.
The main independent variable of interest for our performance analysis is a dichotomous variable capturing whether or not district councils adopted a shared SMT. The use of a dichotomous variable to capture MI is the standard approach taken in the MI literature (see Walker, Chen, and Aravind 2015). We also include the following control variables; first, population figures because local governments serving big populations may benefit from greater purchasing power, and thereby have more resources to improve efficiency and effectiveness (Boyne 1995). Second, a measure of population density to capture the potential to gain efficiencies by providing multiple services from the same site in densely populated areas (Grosskopf and Yaisawamg 1990). Third, deprivation is measured using the average ward score in each district council of the English Index of Multiple Deprivation-the standard measure of socio-economic disadvantage used by the U.K. government, which has been shown to harm performance (Romero, Haubrich, and McLean 2010).
Fourth, measures capturing the age, ethnic, and social class diversity of the population served by district councils are included in the models. The proportions of the various sub-groups within each of the different categories identified by the national censuses within a district's jurisdiction (e.g. ages 0-4, Black African, and higher managerial occupations) was squared and the sum of these squares subtracted from 10,000. These measures capture 'fractionalization' within an area, which has been shown to weaken local government performance (Andrews et al. 2005). Finally, we include lagged dependent variables in our performance models to account for potential auto-regression in local governments' performance (O'Toole and Meier 1999). Descriptive statistics and correlations for all the variables used in our statistical modelling are shown in Table 3.

Antecedents of Shared SMTs
For our quantitative analysis of the antecedents of shared SMTs, we use the panel version of survival analysis, which codes adoption of a shared SMT as a "failure" event-i.e. the cessation of a singleorganization SMT, which occurs on 37 occasions during the study period. The average Variance Inflation Factor (VIF) score for the independent variables in the model is 1.21, suggesting multicollinearity is unlikely to be a serious problem.
The results for the control variables shown in Table 4 indicate that there is a positive but statistically insignificant relationship between the number of companies and the adoption of a shared SMT, a statistically significant positive relationship between financial condition and adoption and an inverted u-shaped relationship for political leader tenure. Moving to our key independent variables, two of the economic antecedents are statistically significant predictors of the adoption of a shared SMT. The hazard ratio for organizational size is less than one and therefore negatively related to adoption of a shared SMT, as per hypothesis 1. The hazard ratio for expenditure reductions is greater than one indicating a positive relationship and supporting hypothesis 3. Although hypothesis 2 proposing that administrative intensity is negatively related to adoption of a shared SMT is not confirmed, the findings for hypothesis 1 and 3 signal the importance of economic motivations.
We explored whether the potential for scale economies was a key motivation for setting up a shared management team with interviewees. Most drew attention to the desire to capture new efficiencies through increased scale. For example, one senior manager emphasized that at first it was not a political decision to share a management team, it 'was about getting that economy scale and getting the savings from being shared'. A typical response is exemplified by a senior manager in another local government who underlined that the search for efficiency, initially 'was around being able to maintain services affordably and around economy of scale and those types of factors'. The objective of reaping scale economies was particularly acute in organizations under fiscal pressure. As one political leader explained, in the wake of budget cuts 'we've gone as small as we can now. What we need to do is grow economies of scale and see if we can build partnerships and build relationships with other councils'. The impact of the cuts on the adoption of a shared SMT was corroborated by another political leader who stated that: 'We didn't realise that government funding would become more troublesome year after year after year and that the cuts to our funding were going to be so drastic. We looked to see how we could make efficiencies that would benefit both authorities'.
The results shown in Table 4 indicate that only one of the political variables is a predictor of the adoption of a shared SMT. The hazard ratio for electoral competition is below one and statistically significant. This finding affirms hypothesis 5 on the way that political risk aversion may determine the adoption of MI by local governments. The complex role that electoral competition played in shaping the decision to adopt or persist with a shared SMT is illustrated by the comments of one manager: 'It was obviously in our case Conservative councils at the time of the alliance, although we've had one small period of… Labour for four years. And actually the election campaign of the Labour group prior to that election was to end the alliance. But within five minutes of getting into power they realised that was going to be pretty difficult to do and so rowed back from it'.
Although hypothesis 6 was not confirmed by our statistical analysis, some of our interviewees emphasized that the timing of elections was a major influence on whether or not to adopt shared management arrangements. Typical of these was a manager who stressed that to push for the adoption of radical MI within local government '…you've got to pick your window for transformation in terms of the political framework'.
We are unable to confirm hypothesis 4 because although the hazard ratio for right-wing political control is above one, it does not achieve statistical significance. Nevertheless, a senior manager indicated that for their partner organizations, the decision to share the SMT 'might have been easier given that they were both Conservative controlled' and a chief executive officer explained, 'there is a political agenda here as well, of course, because it's a very strongly Conservative council'.
The results for the institutional antecedents confirm both hypotheses 7 and 8. There is a statistically significant inverted u-shaped relationship between proximity to central government and adoption of a shared SMT, albeit with a weakly significant base term. There is, however, a strong statistically significant positive relationship between proximity to prior adopters and sharing a SMT. These results provide support for our arguments about the nature of coercive and mimetic institutional pressures on English district councils.
Interviewees drew attention to the coercive role that central government played in driving the adoption of a shared SMT, especially as a potential alternative to forced amalgamation. As suggested by a politician: 'If we didn't do this transformation, didn't go forward with the transformation, that local government transformation [forced amalgamation] will happen, in the future, it's just about when'. In terms of mimesis, there was a sense that district councils were constantly looking to what their neighbors were doing as part of a wider search for new ways of working. For instance, a chief executive officer indicated that: 'We all learn from each other … The arrangements are all different but the basic concepts are the same'. Interestingly, interviewees also drew attention to the normative institutional forces impelling them to adopt a shared SMT, with one senior manager reporting 'It's just good to know that the government is supporting what we're doing and it's the right thing to do'.

Organizational Performance Effects of Shared SMTs
For our quantitative analysis of the shared SMT-performance relationship, we use Arellano and Bond's (1991) Generalized Method of Moments (GMM) procedure, which takes first differences to eliminate individual specific effects, and then instruments potentially endogenous independent variables in the first-differenced equation using levels of the series lagged at least two periods. In addition, we report estimates from a bootstrapped bias correction (BBC) for dynamic panels derived by Everaert and Pozzi (2007). The average VIF score for the independent variables in our performance models is again low at 1.23.
Tables 5 and 6 present the regressions estimating the relationship between sharing a SMT and public service efficiency and effectiveness. Following McQuestin and Drew, McQuestin, and Dollery (2019), we report estimates controlling for a linear (models 1 and 3) and a nonlinear (models 2 and 4) population-performance relationship. To estimate the nonlinear relationship, a squared version of the population variable was added to the models. Since population and population squared do not both achieve statistical significance in models 2 and 4 in tables 5 and 6, we discuss the results from models 1 and 3. In Table 5, the coefficient for sharing a SMT in models 1 and 3 is negative and statistically significant, and in Table 6 the shared SMT coefficient is also negative for models 1 and 3, though it is only statistically significant for model 3. Overall, these results provide support for hypothesis 9, which posits a negative shared SMT-performance relationship.
Sharing local governments devote much of their energies to developing joint management systems, so the transaction costs associated with that process are likely to be high. For instance, one interviewee noted that 'we spent a lot of time re-engineering' management processes. It is also possible that the financial costs associated with transition to a shared SMT reduce efficiency, at least in the short-term. When asked about the impacts of the decision to share their SMTs, interviewees reported that there were significant transitional costs. A senior manager in one local government indicated that: 'the point in 2011 when we went from basically 15 heads of service down to 7, I can remember that the cost of that was over a million pounds in redundancy costs and pension strain costs'. Nevertheless, a chief executive officer in another government noted that the 'savings, I think shared management-around £1.2 million. Annually. Shared across both', which seems likely to help generate efficiency improvements in the long-run. However, in addition to the economic transaction costs of making joint-management systems work, political transaction costs seem likely to have implications for the prospect of performance gains. In particular, the performance management literature suggests that political leaders may choose to "redirect resources to other priorities and willingly accepted the possibility of some slippage" on certain dimensions of performance (Ammons 2013, 523; see also Moynihan 2008). We explore this possibility next.

Discussion
Our findings on the antecedents of the adoption of shared SMTs provide support for the rational and institutional perspectives on MI. Economic rationality appears to be a strong influence on MI adoption, corroborating the idea that fiscal pressures can drive innovation (Singla, Stritch, and Feeney 2018). At the same time, the salience of political and institutional determinants influencing adoption of a shared SMT underline the importance of the distinctive organizational field that we analyze.
Although there has been an explosion of interest in public sector innovation (see De Vries, Bekkers, and Tummers 2015),  comparatively little attention has been paid to the politics of innovation per se, let alone to the politics of MI. Moreover, few studies address the isomorphic pressures on innovation or whether MI results in the performance improvements that its advocates anticipate. By examining MI within local governments, we are able to illustrate the ways in which the political and institutional dimensions of public administration shape the adoption and management of public sector innovation. In particular, the political transaction costs associated with institutional collective action may explain why performance benefits have not emerged. Shared SMTs can find it difficult and time-consuming to reconcile the competing needs and demands of two 'sovereign' political entities in an equitable and effective way.
We undertook further analyses to investigate the political transaction costs associated with sharing in more depth. T-tests comparing pre-sharing performance suggest that shared SMT arrangements involve one high-and one low-performing council: average efficiency score for low-performing partners = .57, average efficiency score for high-performing partners = .67 (t-statistic = 1.87, p < .1); average effectiveness index score for low-performing partners = −.88, average effectiveness index score for high-performing partners = 1.19 (t-statistic = 3.14, p < .01). Following the performance management literature (e.g. Moynihan 2008), this raises the possibility that any joint performance pay-off may be postponed as time and resources are diverted away from the high-performing partner to support their low-performing counterpart. Importantly, it seems that this supportive relationship occurs where two councils share the same political ideology-a finding mirroring research showing how local political homophily facilitates inter-organizational collaboration (Song, Park, and Jung 2018). A quote from a CEO encapsulates this: 'He [a political leader] basically said, "Look we've got a neighbouring council, they were Conservative controlled. We should put an arm around their shoulder and help them out"… One of the dilemmas you have working on a shared arrangement, is that issue around visibility and accessibility and time spent. There is the expectation that you've got to split your time 50:50… You have to be seen at both. So, right from the outset, he (the leader of one council) said, "Look XXX, I appreciate you're going to have a tough job in XXX for the first six months, the year. I understand you won't be able to spend 50% of your time down at XXX. You'll have to spend more than 50% of your time at XXX sorting things out'.
Although they made little mention of tangible improvements in efficiency and effectiveness, interviewees did, however, emphasize that there were positive management outcomes from having a shared SMT: greater resilience and strength in negotiation with central government; a better qualified and motivated workforce; and new skills and capacities. For example, in keeping with resource dependence arguments, a senior manager underlined that 'there was a view that it might actually give us a stronger voice; it gives us a greater powerbase with detractors and so on'. Another pointed toward the attractiveness of working for a larger jointly managed entity 'as we've brought the workforces together, we've retained the very best people. We're able to keep the best people now'-a comment that echoes arguments about the purchasing power of bigger organizations (Black, Noel, and Wang 1999). More generally, there was a feeling that sharing organizations were able to benefit from the strengths that each partner brought to the arrangement, affirming the insights of the collaboration literature (Lasker, Weiss, and Miller 2001). A senior manager highlighted that: 'There's a lot of learning that we're building in from the way that we did it in one authority to apply to the other authority. So, that's one of the most powerful things I've found about it is you learn from each side'.
In sum, local governments sharing SMTs appear to be seeking to build the capacity to respond to challenging economic circumstances and institutional pressures. While this may mean that they struggle to realize gains in efficiency and effectiveness, it could also be the case that local governments with shared SMTs have developed the organizational resilience required to stave off threats to their survival along with the capabilities needed to achieve longer-term performance enhancements. Importantly, though, these potential benefits of joint-management may only occur where local governments are of the same political persuasion. Party-political homophily may reflect an altruistic concern on the part of the senior partner within a shared arrangement or a mutual commitment to certain policy choices (e.g. privatization, cutbacks, or redundancies) that preclude bipartisan co-operation. Further research in local governments adopting shared SMTs and MIs in other countries would cast invaluable light on this complex issue.

Limitations
A mixed method approach has been useful to complement findings on the antecedents and impacts of shared SMTs from a multilevel perspective. Nevertheless, the study has a number of limitations that create opportunities for subsequent investigation.
Firstly, our analysis has examined a specific MI in a particular group of public organizations during a specific time period. It is therefore important for subsequent studies to explore whether the antecedents and effects of shared SMTs and other MIs differ in other time periods and organizational settings. In particular, sharing a SMT may be more common in England than elsewhere due to the greater influence of central government on local governments in the country. Comparative analysis of the antecedents and effects of shared SMTs and related structural reforms across multiple countries would therefore add greatly to our understanding of the best approaches to reforming local government systems. Secondly, the institutional collective action framework points towards the role of service characteristics in shaping the successful introduction of structural innovations (Hansen, Mullin, and Riggs 2020). It would therefore be useful to investigate whether there may be different models of sharing a SMT in place in different types of local government (e.g. single purpose versus multi-purpose governments), and the implications of these different models. For example, cases of SMT sharing have been observed in London borough councils in England, local governments that are responsible for a much wider range of services than district councils. Thirdly, future research should utilize quantitative data to investigate the effects of shared SMTs on other dimensions of performance, especially citizen satisfaction, for which longitudinal data for English local government was not available for our study period, and the resilience and strategic capacity that were identified as important effects in our qualitative interviews. Finally, while we provide evidence of the salience of economic, political, and institutional antecedents for the adoption of MI by local governments, other relevant variables could potentially be incorporated in subsequent studies. For example, the leadership style of CEOs and politicians are arguably important determinants of public sector MI (Crosby, t'Hart, and Torfing 2017;Mulgan and Albury 2003). These too may play a part in shaping the performance effects of shared SMTs and other MIs.

Conclusion
This article suggests that public sector MI may be driven by political and institutional factors as much as economic circumstances. The salience of institutional pressures for sharing a SMT indicate that MI is only adopted when public organizations are forced to change or do so to keep up with their neighbors-something that may account for the apparent absence of performance benefits. Likewise, the distinctive political influences on MI may explain why, in this instance, efficiency, and effectiveness did not improve. If MIs are introduced when the risks to incumbents' re-election are low or to bring the performance of neighboring organizations up to a certain level, then they may not be occurring at the optimal time for generating subsequent improvements. These findings contribute to theories of innovation in public organizations and can assist scholars and practitioners in further understanding the motivations for adopting MI in the public sector and the effects that this may have on performance.